The invention relates to the field of
plant phenotype recognition and intelligent breeding, and discloses a multi-
source data fused cotton
phenotype characterization and breeding decision
convolutional neural network platform, which comprises an
image acquisition module, an environment sensing module, a
molecular marker input interface, a data preprocessing module, a multi-
modal feature extraction network, an attention fusion module and an
intelligent decision engine. Image features are extracted through a
convolutional neural network, environment response features are modeled through a gated
loop network, a uniform phenotypic vector is generated in combination with
molecular marker embedding, multi-source features are weighted and fused by adopting an attention mechanism, and high-dimensional phenotypic representation is constructed; and based on the weighted
cosine similarity between the target
character vector and the candidate individual vector, outputting a sorting result and
mating combination recommendation. The method can realize multi-factor joint modeling of complex agronomic traits and target-oriented breeding path decision, has the advantages of high accuracy, high
interpretability and high decision transparency, and is suitable for precise breeding and intelligent recommendation of large-scale cotton materials.